Skip to content

πŸ§‘πŸ»β€πŸ’» Multivariate Time Series Modelling with Neural SDE driven by Jump Diffusion

License

Notifications You must be signed in to change notification settings

kirillzx/JDFlow

Folders and files

NameName
Last commit message
Last commit date

Latest commit

Β 

History

94 Commits
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 

Repository files navigation

JDFlow

Kirill Zakharov

The official realisation of the prposed method JDFlow at the conference ICCS 2024, Spain, Malaga (link on the article will be available later).

Abstract:

Training hyperparameters

We employ three methods for generating multivariate time series to compare against our approach: Fourier Flow, fractional-SDE and PAR from SDV. We train Fourier Flow for 1000 epochs with 10 flows. The hidden dimension of the model is equal to 256. For the PAR synthesizer we utilise $1000$ epochs and for a metadata we additionally use the time index with uniform range. The fractional SDE-Net model was trained by 1000 epochs and batch dimension 5.

About

πŸ§‘πŸ»β€πŸ’» Multivariate Time Series Modelling with Neural SDE driven by Jump Diffusion

Topics

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published